Myth Busters #3: Hysterectomies in Lewiston, Maine
One of the consequences of Roemer’s Law has been the idea of “provider induced demand,” and the general notion that everything that happens in health care is because some greedy doctor has deemed it. This means that patients don’t count. What patients may want is irrelevant.
Nowhere is this better illustrated than in Jack Wennberg’s early work on “small area variation” in medical practice.
I was working in the research department at Blue Cross Blue Shield of Maine from 1979 to 1984 and we offered the use of our claims files for his research. He had already done some work in Vermont looking at variations in the rate of tonsillectomies in various towns. He found that in some places physicians surgically remove tonsils at a much greater rate than they do in other places. He concluded that this was an example of Roemer’s Law in effect — scalpel-happy physicians were too quick to order up surgery in some places, but not in others.
The most startling variation he found in Maine — and the one that put him on the map — was the difference between the rate of hysterectomies in Lewiston and Wiscasset, just 35 miles away. The chance of a woman getting a hysterectomy in Lewiston by the time she was 70-years old was 70%. In Wiscasset it was a fraction of that.
Wennberg, in keeping with the narrative he developed in Vermont, decided that this was because physicians in Lewiston were eager to cut while their fellows in Wiscasset were not. Curious. Why should that be? He didn’t know but speculated that maybe they were trained at different medical schools, or somehow the profession in one town had grown to be more aggressive than those 35 miles away.
Hmm, is Maine so very isolated that doctors in Lewiston never talk to their colleagues just 35 miles away? That was certainly not my experience. In fact, Maine is not exactly crawling with physicians, so opportunities for professional bonding would necessarily include physicians from many different towns. In fact, in a recent search of a physician directory compiled by the Maine Medical Center, I could find only 45 Obstetricians/Gynecologists in the entire state. Is it really credible that those only 35 miles apart would not be talking with each other?
Completely missed in Wennberg’s analysis was the possibility that maybe it wasn’t the physicians who dictated what happened, but the patients. There are stark differences in the populations of the two towns. Wiscasset was an old fishing village with a largely Yankee (Protestant) population. Lewiston was an industrial mill town with a predominantly French-Canadian (Roman Catholic) population. It is far more likely that women in Lewiston were using hysterectomies as a form of birth control that was acceptable to the Church, especially in the early 1970s before birth control pills were widely available. Many of us who lived in Maine at the time, thought this was pretty obvious, and were surprised and amused at the attention Wennberg’s study got.
Even more surprising is that Wennberg’s study is still getting attention thirty years later. In 2009, National Public Radio (NPR) published a major article, “The Telltale Wombs of Lewiston, Maine.” The piece by Alix Spiegel is gushing about Wennberg, calling him, “a certified guru — a man whose insights underlie many of the arguments you currently hear about health reform….” It adds, “Over the past 40 years, he has completely transformed our understanding of what’s going on in the U.S. health care system.”
The article goes on to say there were two possible explanations for differences in medical practice — “The first explanation was that doctor behavior was somehow to blame. The second explanation was that it was the patients; that people in some areas were just much sicker than people in other areas, or maybe just wanted more services for some reason.” The article says Wennberg looked at the possibility of differences in patient demographics and level of sickness and dismissed it.
This is unconvincing because it doesn’t say whether he gave any weight to issues like religious differences rather than just differences in disease severity. It is a profound weakness in research that treats patients like statistics rather than fully developed human beings. Statistics fail to capture the most important aspects of human beings, such as personal values or emotional condition. This is precisely why medicine has always (until now) been centered on a personal relationship between a doctor and a patient.
Wennberg insisted that, “it was doctors, not patients, who drove medical consumption, and all kinds of things influenced the decisions a doctor makes when a patient enters his office.”
Actually, this reveals a bias among the educated elite that “common” working people are not bright enough or involved enough to have much effect on the world around them. An academic like Jack Wennberg assumes that all decisions must be coming from people like himself — other educated elitists. “The people” are just the raw material the elite uses in their machinations. So, to the extent “the people” have opinions, preferences, fears, hopes, values, or expectations, these are just characteristics that have to be managed by the people who know best.
After much discussion, the NPR article finally gets to the heart of the matter — money. It says, “the truth is the decisions made by your physician when you enter his office are profoundly influenced by the way that doctors get paid in this country.” It quotes Gordon Smith, the non-physician head of the Maine Medical Association, as saying, “If you pay people more, the more things they do, they’re going to do more things.”
That is the conventional thinking among academics these days, and it goes back, once again, to Roemer’s Law (“a built bed is a filled bed”) and the idea of “provider induced demand.” Once again, that notion is driving most of the thinking in health policy today.
But Wennberg’s research, if anything, contradicts that very idea. It does not support it. Physicians in Wiscasset and Lewiston were all under the same payment system. They were all subject to the same incentives. If it were the payment system that drove their decision-making, they would all be practicing in the same way. The fact that they practice differently strongly suggests it is not the payment system that drives behavior.
But once again public policy is driven by an idea that could fit on a bumper sticker — “Greedy Doctors rip out wombs for fun and profit” — even though the doctors in two towns 35 miles apart behaved very differently.
Still, Wennberg went on to create the Dartmouth Health Atlas, which lives on and continues to distort data and influence public policy, all based on this original idea that the payment system incentivizes greedy doctors to over treat hapless patients — but only in some towns.
Very interesting. Thanks for the post, Greg.
Something that is often lost in the econometric regressions is the cultural elements that were never captured in the dataset used. Researchers make inferences without really ever knowing the actual facts.
We have seen the same dynamic at work here in Grand Junction, touted as the ideal model during the Obamacare debate. Dartmouth Atlas indicated Grand Jct expenditures were lower than other places. The local Medical Society trumpeted that this was due to their model of almost monopolistic managed care, totally ignoring the geographic and demographic aspects of the population as well as ignoring differences in distribution of physician specialties compared to other areas. To me it seems like cramming the data to fit an agenda rather than letting the data tell the real story.
Nice piece of analysis, and great example, Greg! That said, there is the “patient freebie goodies” incentive as well. Nobody is going to get a hysterectomy without a good reason, medical or otherwise. But lots of people will get a second MRI when one will do, an extra specialist visit each week (a social event) when one will do, etc. Why get the plain Vanilla pacemaker when the platinum-plated one is free. So there is lots of “blame” to share around.
Interesting post Greg. As someone who has read a lot of Dr. Wennberg’s work (and who has a different viewpoint than the majority of the authors/commenters on this blog) I would take issue with a number of your assertions. I do definitely enjoy reading criticisms though.
1. Regarding the disparity in hysterectomies between Lewiston and Wiscasset, Wennberg did not dismiss the role of religion as explaining part of the disparity, but rather, indicated that it could not explain away the entire disparity. The religious difference (and potential differences in patient demand) does not necessarily mean that differences in physician practice did not also play a role. Barring data (as opposed to anecdotal evidence) that shows that patient demand can explain the entire variation, it seems premature to dismiss alternative explanations.
2. Wennberg’s theory that differences in physician practice (rather than patient demand) was influencing the surgery and hospitalization rates was based on a series of findings in a number of regions for a number of surgeries. Even if you insist that religion-driven patient demand is the only explanation for the difference in hysterectomies between the two Maine towns, I’m not sure how demographic or religious differences would necessarily influence the rates of other surgeries like back surgery, tonsillectomy, etc. (if you have a theory, please enlighten me, preferably one supported by data). These disparities were also observed between localities that were nearly identical demographically.
3. Regarding the simplistic payment system critique (if two towns have the same payment system, how could they have variations not explained by patient demand?), I think you are mis-characterizing or perhaps misunderstanding Wennberg’s hypotheses. I think his idea is that the payment system exacerbates underlying issues such as differences in physician supply, supply of hospital beds, and local physician culture (of course, if you reject the extensive data supporting the idea of supply-sensitive care, then we simply may never see eye-to-eye).
4. As for the idea that Wennberg is too elitist and educated to think that the concerns and values of the common people matter, how would you explain his early and consistent advocacy of shared decision making (SDM)? The whole principle of SDM is that exposing patients to the available evidence supporting their treatment options, and consulting the patient about their wishes and values, will help the patient make the best medical decision (and research has documented that patients tend to be more satisfied with their treatment following SDM). The whole purpose of SDM is patient empowerment, and helping patients be able to make more informed decisions. How does this advocacy fit in with your ideologically-driven narrative of Wennberg as a heartless, elitist academic who wants to dictate behavior from his ivory tower?
I’m by no means an ideological Wennberg supporter, and I do have concerns about drawing overly broad conclusions about the Dartmouth Atlas data. However, your critique of his motives (which comes off as ideological and uninformed) diminishes your critique of his methods (which has merits, but as I argued, is not sufficient to debunk his conclusions, which are based on pretty comprehensive research).
I’d enjoy the chance to discuss this further with you, and would be curious about hearing your (or somebody else’s) response to my points.
Eric,
I appreciate the comments. Unfortunately, we are doomed to critiques based on anecdotes until The RWJ Foundation or some other group will fund some original research. But in science an anecdote can indeed disprove a theory. If the theory is that water freezes at 32 degrees F and I can find a case where it does not, the theory is disproven.
In this case, Wennberg has offered zero evidence of in what way physicians are so different from each other that they choose to practice differently. Did they go to different medical schools? Are they from different cultures? Is one batch female while the other batch is male? Is one batch left handed while the other is right handed? What exactly is there about one group of MDs that makes them different from another group? I haven’t even seen any conjecture on it, let alone sufficient evidence to drive public policy.
For another example of Dartmouth completely ignoring patient differences see my post on end-of-life care — There s a pattern here that is profoundly myopic.
Eric made the point I would have made. You. Greg, are, perhaps, more eager than Wennberg to jump to a demand side, vs supply side, explanation.
That said, I enjoyed reading your post and learning of your alternative explanation.
My thought on the Wennberg approach has always been that you don’t look for mere variation, but unexplained variation, unexplained by demographics certainly. One idea might be that the religious culture of the consumers was magnified by the possible religious orientation of the physicians.
This is a great piece Greg. What we should take away from it is the realization that just about every contentious subject these days (including healthcare) is really down to a single ideological battle. All the statistics engendered in the fights, both local and national, are only tools to be used (often incorrectly) to prove an argument.
It really all boils down to the Marxist/Liberal/Statist/Elitist school of thought versus the Free Market/Libertarian/Independence minded school of thought. The information made readily available through digital technology has drawn sharper definition between the ideologies, and it seems safe to predict that most of those in the squishy middle will be forced to actually think through the process (which they have never really had to do before), and make a decision about their philosophy of life. Being “just a little bit pregnant” doesn’t seem to make sense anymore.
A good way of coming to a conclusion about the reasons for hysterectomies would be to look at the ages at which women underwent them. The post notes that the chance of hysterectomy was 70% by the time a woman is 70, but it doesn’t say anything about a woman who is 40.
By comparing the distributions of ages at which women receive hysterectomies between the two towns (and possibly the US/state distributions), you would probably get a good picture of whether or not religious views came into play. I daresay that a 55-year-old woman would not seek a hysterectomy as birth control, but a 40-year-old might.
Frank, it’s that kind of attitude that prevents real data-driven reforms from being implemented to improve health care. I agree that we should be critical of statistics presented, particularly from people with obvious ideological agendas, but that doesn’t mean that the data are inherently invalid.
To just stick your fingers in your ears and whine about bias every time statistics are presented that may disagree with your ideological perspective does a disservice to real debates about how to fix the health care system. Not to mention referring to your ideological opponents as “Marxist”, which makes it hard to take any argument seriously.
John,
Great point. My point was mainly that patient preference and physician practice patterns are not mutually exclusive causes of variation in surgery and hospitalization rates. Bringing up the idea of patient preference as a factor does not automatically rule out the influence of physicians.
Greg,
I agree that the underlying characteristics that distinguish physicians with different practice patterns may be difficult or impossible to measure (and have not been documented, to my knowledge). But once again, that does not rule out the physician influence. What is known is that significant variation exists after adjusting for many of the meaningful differences between patient populations, and differences in patient demand have not proven sufficient to narrow the gap.
Eric (and John),
I’m not saying physician influence makes no difference at all. What I am saying is that Wennberg, Dartmouth, Fisher, and most of the movers and shakers in health policy have all but ruled out any patient influence, and their conclusions are based on the most paltry evidence. There is nothing — NOTHING — to suggest that physicians in Lewiston are any different than physicians in Wiscasset. But there is PLENTY to suggest profound differences in patients. Ditto with end-of-life care in New York City versus Ogden, Utah.
Three possible causes have been advanced for the variations —
1. The payment system. But the payment system is identical.
2. Physician preferences. But there is no evidence that the physicians are different.
3. Patient preferences. There is plenty of evidence that patients are quite different, at least if we go beyond disease severity measures to include socio/demographic differences.
To smugly settle on physicians as the cause is simply dishonest and should discredit the work of these researchers.
Did anyone ever check to see if one doctor was referring patients to the other town because that one was better at the procedure? Was the town census based on where the procedure was done or the location of the patients home?
@Eric
“Frank, it’s that kind of attitude that prevents real data-driven reforms from being implemented to improve health care. I agree that we should be critical of statistics presented, particularly from people with obvious ideological agendas, but that doesn’t mean that the data are inherently invalid.”
Eric, that was not so much an “attitude” as it was (is) an observation. We have been inundated with garbage statistics from day one (for example – American/European birth mortality data) which in part has been used to justify Obamacare. The public cannot rely on statistical evaluation published from sources such as NPR or the Commonwealth Fund. Oh it’s all well and fine that those of us who are somehow involved in the healthcare industry professionally can fight for the high ground of intellectual righteousness, but in the meantime the public is being deceived with lies and half truths using what the media honors as statistics.
So yes, ideology is moving ahead, not guided by factual statistical data, but the other way around. That observation is not gathered from sticking fingers in ears. And perhaps it would be helpful if you researched the political history of some of the presidential staff, advisers, mentors, etc. Maybe you could come up with different and more er…pleasant political labels. I can’t.
@ everybody
I read this thread with amusement yesterday. Then this morning as the sun was coming up it dawned on me (heh) that the posts on this thread collectively present the best argument I’ve ever heard against the federal government being involved in the practice of medicine or so-called health insurance. If several reasonably intelligent and focused people cannot determine why some women in two small towns in Maine have uteruses and some do not, how can anybody expect Kathleen Sebelius and her crew at HHS to come up with a comprehensive set of rules and regulations that are “settled science” or “best practice” for the complex day in and day out medical decisions of 300 million people with all the varied nuances of rational thought, ideology, superstition, religion and every other factor. We all (physicians included) are trying to find our way through a complex world with a mix of knowledge, belief, experience, education, hopes and desires. Thank you all for pointing this by tying yourselves into knots on one simple little example!!
Amazing things here. I am very satisfied to peer your post. Thank you so much and I am having a look ahead to contact you. Will you kindly drop me a mail?